This invention relates to techniques for fabricating patterned poled structures and devices from polable materials, and in particular periodically poled dielectric structures, as well as devices incorporating periodically-poled structures. The invention has wide application in optical systems, and in particular to nonlinear optical systems, such as frequency doublers, optical parametric oscillators and amplifiers, optical waveguides and the like.
The ferroelectric materials of choice for nonlinear optical interactions are LiNbO.sub.3, LiTaO.sub.3 and KTP. Ferroelectric domain inversion is generally produced by in-diffusion and heat treatment, e.g., Ti in-diffusion, proton exchange and Ba in-diffusion respectively for the three materials. In LiNbO.sub.3 and LiTaO.sub.3 these poling techniques generate shallow domains in the polable materials, only suitable for shallow waveguide interactions. For maximum overlap with the optical waveguide modes and optimum device efficiency the domain inverted grating should have a rectangular shape with vertical domain walls. However, in LiNbO.sub.3 the titanium in-diffusion process (U.S. Pat. No. 5,036,220) creates a triangularly shaped domain inversion grating and in LiTaO.sub.3, proton exchange and heat treatment at around 600.degree. C. (U.S. Pat. No. 5,249,191) produces semicircular shaped domains. These odd shaped, non-rectangular domains reduce the efficiency of the desired nonlinear interaction.
In KTP the inverted domains generated by Ba diffusion (U.S. Pat. No. 5,157,754, Van der Poel et al Appl. Phys. Lett. 57 (20), p2074-6, 1990) are rectangular with vertical sidewalls due to the anisotropic nature of the diffusion process. But they are relatively shallow: they can be used for waveguide interactions but are not very efficient for bulk applications. An added complication to the KTP domain inversion process is that the domain inversion is performed in the same ion-exchange step as is used to form the optical waveguide. Although this means that only a single processing step is required to fabricate the device (rather than separate domain inversion and waveguide fabrication processes), it also means that the waveguide and domain inversion dimensions cannot be in-dependently optimized to achieve maximum performance. In order to obtain a periodically inverted domain pattern a segmented waveguide is formed consisting of alternating guiding ion-exchanged regions and non-guiding unexchanged crystal. The application of a different domain inversion procedure, such as electric field induced poling, would separate the two fabrication processes, allowing independent optimization.
The domain inversion processes available in the prior art are limited to the fabrication of relatively shallow surface features that are only suitable for waveguide based interactions. [With the exception of periodically poled fibers of LiNbO.sub.3 grown by the laser heated pedestal growth technique as described by Magel et al (Magel et al Appl. Phys. Lett. 56 (2), p108-10, 1990). However, these periodically poled fibers suffer from the inability to control the domain dimensions and boundary positions to a sufficient degree to ensure phase matching over a long sample length greater than about 3 mm.] For higher power interactions, such as the frequency conversion of high power and pulsed laser systems, waveguide devices are unsuitable because the intensity generated on the front face of the crystal when launching into the waveguide far exceeds the material damage threshold. To avoid physical damage to the nonlinear crystal the high power beam focus must be relaxed. For periodic poled material to be useful in such cases, the poling must be quite uniform across the aperture of the bulk crystal. A domain inversion process is needed which will provide bulk periodic poling in ferroelectric materials.
Electric field-induced poling, as demonstrated by Yamada et al in LiNbO.sub.3 (U.S. Pat. No. 5,193,023), offers the possibility of vertical walled domain structures for maximum performance and deep poling suitable for quasi-phase-matching of bulk interactions in all ferroelectrics. Yamada et al describe the use of patterned metal electrodes 22 deposited on the crystal .+-.z surfaces combined with the application of a pulsed high voltage 20, as illustrated in FIG. 2a. The high voltage pulse exceeds the crystal coercive field and produces domain inversion 12 under the metal electrode. However, frequent destructive electrical breakdown events remain a problem and have so far prevented this poling technique from being adopted for device production. Reports of electric field poling work in the scientific literature [Yamada et al, Appl. Phys. Lett., 62 (5), p435-6, 1993, Yamada et al, ASSL and Blue Green Lasers 1993, paper CThA1-1, Optical Society of America.] are limited to crystal substrate thicknesses of 100 .mu.m due to the destructive effects of electrical breakdown. Larger apertures are required for bulk crystal applications.
Following the approach of Yamada et al, Burns et al [Burns et al, Paper CThC3-1, Compact Blue Green lasers 1994, Optical Society of America, Burns et al, Phot. Tech. Lett., Vol 6, p252-4, 1994] have very recently achieved domain inverted structures in about 250 .mu.m thick LiNbO.sub.3 slabs. However, electrical breakdown events remain a problem, and the width of the inverted domains is not well defined by the photolithographically patterned metal electrode, significantly reducing the efficiency of the device.
The use of electric fields to pole ferroelectric crystals in a patterned form for quasi-phase-matching was proposed in 1980 by Papuchon et al (U.S. Pat. No. 4,236,785) based on x or y-cut crystal substrates. In that invention, crenulated electrode structures 24 were fabricated by photolithography on the x or y-cut surface 26 of the optical crystal as shown schematically in FIG. 2b, with a period defined by the coherence length of the desired nonlinear interaction. A high voltage is applied across the two electrodes such that the electric field is antiparallel to the crystal z direction (the spontaneous polarization direction of the material) to periodically domain invert the crystal substrate. This periodic poling technique has since been realized using LiNbO.sub.3 as the crystal substrate, and reported in the scientific literature, [Janzen et al, Integrated Photonics Research, paper TuD5-1, 1992, Optical society of America], but the inverted domains are necessarily very shallow (.ltoreq.2 .mu.m) because the electric field between the coplanar electrodes does not penetrate deeply into the crystal.
Also proposed by Papuchon et al was the deposition and patterning of a second electrode over that used for domain inversion, allowing the application of different voltages to the inverted and uninverted domain sections in the quasi-phase-matched grating, and the subsequent tuning of the quasi-phase-matched interaction wavelength peak using the electro-optic effect.
Matsumoto et al (Electron. Lett., 27, p 2040-2, 1991) describes the second harmonic generation of blue light in electrically periodically poled LiTaO.sub.3 waveguides. The z-cut LiTaO.sub.3 crystal substrate was poled by applying a periodic electric field using interdigital electrodes while heating the sample to just below the Curie temperature about 610.degree. C. The conversion efficiency of the waveguide device fabricated was significantly lower than those fabricated by proton exchange on the +z face and has not been pursued further.
Siebert et al (Proc. Soc. Phot. Instr. Eng., 1362, p370-6, 1991) report ferroelectric microdomain reversal on y-cut LiNbO.sub.3 surfaces using the pyroelectric effect to generate an electric field. A pair of crenulated electrodes is patterned on the y-cut face separated by about 6 .mu.m along the z axis and with a period of 30 .mu.m. Heating the crystal to about 100.degree.C. generates sufficient voltage across the electrodes, due to the pyroelectric effect, to spatially periodically invert the ferroelectric domains between the electrodes. The electric fields do not however dip deep into the crystal, so the resulting domains are only about 1 .mu.m deep, too shallow even for waveguide based interactions to be efficient.
Mizuuchi et al (Appl. Phys. Lett., 62, p1860-2, 1993) disclose a technique for fabricating periodic domain inversion in x-cut LiTaO.sub.3 using selective proton exchange and quick heat treatment, as performed more commonly on the z-cut face. The domain inverted regions are inclined at an angle to the z-axis in the y-z plane. However, quasi-phase-matched nonlinear interactions have yet to be demonstrated in such a domain inverted structure.
Electric field poling is not limited to ferroelectric materials. Brueck et al (U.S. Pat. No. 5,239,407) report the creation of a large second order nonlinearity in fused silica when exposed to a high electric field (about 5 kV/mm) at elevated temperatures (250.degree.-325.degree. C.). The stable nonlinearity is thought to be due to the migration of ionic contaminant species within the fused silica under the influence of the applied electric field.
Organic polymer materials can also be poled to produce second order optical nonlinearities by the application of electric fields (e.g. U.S. Pat. No. 5,064,265). A pendant group which exhibits second order nonlinear susceptibility is added as a side chain to a thermoplastic polymer. Poling is achieved by heating the polymer near or above its glass transition temperature, applying a DC electric field to align the molecular dipoles of the side chain pendant groups in a uniaxial orientation, and cooling the polymer while still under the influence of the electric field to fix the alignment of the side chain molecules.
The current art is unable to provide a number of features required for the manufacture of cheap visible light sources, based on quasi-phase-matched second-harmonic-generation, and bulk periodically poled material for nonlinear optical interactions. These problems are poor reproducibility from sample to sample, shallow inverted domains unsuitable for bulk optical interactions, and the destructive effects of electrical breakdown on the crystal substrate.
The field of integrated optic waveguide frequency conversion devices has, in the prior art, been based almost exclusively on z-cut ferroelectric materials because the best developed of the known domain inversion processes are only applicable to z-cut surfaces and only produce shallow inverted domains. This is the case with titanium in-diffusion into LiNbO.sub.3 and barium ion exchange in KTP. To access the largest nonlinear optical coefficient in these materials, d.sub.33, the pump and signal beams must be polarized along the z axis, requiring a transverse magnetic (TM) polarized waveguide mode (polarized perpendicular to the plane of the waveguide chip). However, semiconductor diode lasers, which are the most attractive pump sources for many nonlinear optical frequency conversion interactions, are polarized in the transverse electric mode (TE), i.e. polarized parallel to the plane of the diode, and therefore require a costly polarization rotation element to be inserted between the diode and nonlinear crystal. There is significant need therefore for a manufacturable technique of periodically poling x or y-cut ferroelectric crystals so that the waveguide mode can be TE polarized to match the diode laser input and still access the large d.sub.33 nonlinear optical coefficient.
Electric field poling is difficult since the electric field strength required for poling is close to the threshold for electric breakdown. In order to achieve poling, the applied field must be above that required for poling, but not too far above the breakdown field. In practice, this is very difficult to achieve, and the difficulty becomes more severe with larger samples where the breakdown field can vary across the sample.
The prior art has an intrinsic difficulty with breakdown. Due to the electrode configuration, portions of the material experience an electric field which is significantly higher than the average applied field in the bulk of the sample. The electrode pattern which is used in the prior art concentrates the electric field, resulting in severe breakdown problems.
It has been determined that the problem in the prior art is caused by the electrode geometry, which is typically formed photolithographically into a pattern as shown in FIG. 2a (Prior Art). A metal film 22 is patterned to form a continuous contact layer over the regions 12 where the poling is desired, and the film is removed over the areas 13 where the original poling is not to be disturbed. A field is applied between electrodes 21 and 22, typically during a short pulse, and poling is achieved above a certain applied voltage. Unfortunately, in many of the samples electric breakdown occurs at some time during the applied pulse, and the resulting arc discharge physically destroys the sample. Breakdown occurs with such a high probability that electric field poling has not been considered as a viable method for producing domain-inverted ferroelectric material.
The problem with the prior art electrode arrangement can be seen from an analysis of the electric field pattern at the surface of the substrate material. In the geometry of FIG. 2a, the normal component of the electric field on the surface in the gaps 23 between electrodes is almost exactly 0.5 times the value E.sub.o of the electric field in the bulk far from the patterned electrode. This is shown for example by J. D. Jackson in his well known book Classical Electrodynamics (Second Edition), in the section treating mixed boundary conditions. From Gauss' Law, it follows that the average normal field along the surface under the electrodes must be substantially higher than the bulk field E.sub.o. In LiNbO.sub.3, it has not been possible to consistently apply a field E.sub.o which exceeds the poling threshold without also exceeding the breakdown threshold somewhere in the sample. This problem is due to the enhancement of the field under the electrodes.
If the spatial duty factor of the electrodes is 50% (i.e., the width of the electrodes is 50% of the period), the average normal field under the electrodes is 1.5.times.E.sub.o. The peak field under the electrode is even larger. From Faraday's Law, we know that the normal component of the electric field varies continuously as a function of the position x from its value of 0.5.times.E.sub.o at one edge of the electrode up to its peak value and back down to 0.5.times.E.sub.o at the other edge of the electrode. We can estimate the magnitude of the peak field even without an explicit solution of the electrostatics problem if we assume a reasonable functional form for the x-variation of the normal field. Assuming a sinusoidal dependence, one can calculate a peak normal field of 2.5.times.E.sub.o. In reality, the peak field is somewhere in the neighborhood of twice the bulk field.
The room temperature breakdown field of LiNbO.sub.3 is reported by the Sony group (Yamada et al) to be around 26 kV/mm. The required field for poling is between 22 and 24 kV/mm at room temperature, which leaves a margin of no more than 20%. The bulk field must exceed the poling field or domain reversal will not propagate through the sample. (While it is known that the poling field decreases at higher temperatures, the breakdown field also decreases, so that no large advantage is obtained at high temperature.) To achieve poling, the applied field must exceed the poling field, but the applied field must not exceed the breakdown field by too much for too long, or the sample will be destroyed by a catastrophic electron avalanche. The margin between the required and the maximum fields is clearly narrower than the factor-of-four variation in the applied field.
The simple periodic electrode structure practiced in the prior art cannot maintain the largest (peak) applied field below the breakdown field in LiNbO.sub.3 while still exceeding the poling field in the bulk of the polable material 8. A similar problem exists in other ferroelectric materials such as KTP and LiTaO.sub.3. A technique is needed to allow periodic poling while not exceeding the breakdown field. A periodic electrode structure is needed which modulates the applied field above and below the poling field along the surface of the sample (to control the spatial extent of the regions which are to invert their nonlinear optical polarity), while controlling the peak field so that it does not exceed the breakdown field by too much or for too long during the applied voltage pulse.